The Sameness List: A Single-Entry Book
On why “we already covered this” produces no discrepancy when it is wrong
The most frequent resource-allocation decision in language teaching is “we already covered this.” Once it holds, the item receives review treatment: minutes instead of a lesson segment, a nod instead of independent production. On the proposition that identity depends on a purpose-relative list of properties, this paper is not the first arrival: that similarity is an incomplete predicate requiring “in what respects” is a 1972 philosophical result (Goodman); that categories are derived from current goals is a 1983 experimental one (Barsalou); that classification systems are built, value-laden and consequential is a 1999 monograph (Bowker & Star); and that “word families and lemmas” is not a real dilemma because the right unit depends on purpose was published in this field’s flagship journal in 2021 (Webb). On the proposition that teachers and learners hold two non-coinciding categorization standards, the 1981 expert–novice sorting study said it all, and the classroom version had data by 1989. This paper therefore claims neither. It claims the position they jointly leave empty: in teaching, this list is kept by single-entry bookkeeping. It records only the teacher’s side, has no contra account, and converts directly into time and checking intensity; a wrong judgement therefore produces no discrepancy anywhere — unlike double-entry, nothing fails to balance — and so it is never discovered.
sameness judgement; single-entry bookkeeping; unreconciled items; purpose-relative grain size; allocation trigger; invisible discrepancy
The Sampling Frame: The Candidate Set Is Generated by the Conditions, Not Given Before Observation
All three source fields assume a pre-given candidate set; taphonomy’s own material overturns it
A learner answers one item correctly and is said to have mastered the point. The usual worry is coverage, or how many more items are needed. This paper claims neither. That more measurement raises stability without narrowing the coverage gap, that which facet to add depends on its variance component, that the information a trial yields depends on the current prior, and that the same evidence means different things under different nomological nets are all prior art — 1959, 1956, 1972, 1955. The paper therefore does not claim that varying conditions beats repeating: that is occupied, verbatim, since 1972. It claims one thing: in teaching, the set of competing explanations is not given before observation; it is generated by the conditions of observation. Hence the primary work of changing conditions is not to discriminate better among live candidates but to let a class of explanation become a candidate for the first time. The overturning material is internal to taphonomy: with fossils, the sampling and the biases were done by natural processes before research began, and a tradition made wholly of perishable material appears not as missing but as nonexistent. The countable reading is therefore not coverage but the novel-candidate rate — the share of observations whose result falls outside every explanation listed beforehand — predicted to approach zero under repetition and to be clearly positive under changed conditions.
candidate set; generated by conditions; novel-candidate rate; taphonomic bias; likelihood ratio; pretest state
Where One Can Get a Question In
The longest same-side run as a reading of a stalled turn-taking loop
Language teachers report a puzzle: the better prepared and more fluent the lesson, the fewer questions. On “asking requires an identifiable starting point,” this paper is not the first arrival — that thinking begins in perplexity and a forked-road situation is a 1910 claim (Dewey); Uncertainty and epistemic curiosity is a 1962 title (Berlyne); To ask a question, one must know enough to know what is not known is a 1979 title (Miyake & Norman); Anomalous information triggers questions is a 1993 title (Graesser & McMahen). On “fluency is misread as good teaching,” a 2013 experiment already showed that instructor fluency raises perceived learning without raising actual learning (Carpenter et al.). On “this is a reciprocal, write-back loop,” conversation analysis (adjacency pairs, repair as a self-righting mechanism), the interaction hypothesis (negotiation of meaning and modified input), niche construction (reciprocal causation) and two-sided markets (the chicken-and-egg problem) each already supply a complete model. The paper claims none of the three. It claims the one reading they jointly leave empty: arrange a lesson’s initiating moves in time; the longest same-side run is a reading that the loop has stalled. Once stalled, incentives applied to the non-initiating side do nothing — the same result two-sided markets reached: subsidise the other side.
question site; initiation; longest same-side run; stalled loop; two-sided condition; supply of starting points
Filling Is Closing: The Help Itself Is the Depleting Event
A claim about an autophagic structure, and how it differs from expertise reversal, cognitive reserve, and degeneracy
When a language teacher supplies a correct form to a stuck learner, every current ledger records a gain. All four are additive ledgers: they record what was added, none records what was closed. On the proposition that help closes a route, this paper is not the first arrival. Seventeen prior claimants approach from five directions — reduction in degrees of freedom (Wood, Bruner & Ross 1976), preference for self-correction (Schegloff, Jefferson & Sacks 1977), wait-time (Rowe 1974), blocking (Kamin 1969), aptitude-treatment interaction (Cronbach & Snow 1977), expertise reversal (Kalyuga et al. 2003), degeneracy (Edelman & Gally 2001), cognitive reserve (Stern 2002) — five of them before 1980. This paper therefore does not claim that help has a cost, nor that the cost depends on a learner-side remainder; both are taken. It claims the one structure that is not taken: the act of helping is itself the depleting event. In aptitude-treatment interaction the moderator is exogenous; in expertise reversal the moderator is cultivated by instruction; in degeneracy and cognitive reserve the depleting events are pathology and lesion. Only in the classroom are the event that supplies the help and the event that consumes the remainder the same event. An intervention consumes its own moderator: an autophagic structure. The material that overturns the shared assumption comes from dentistry itself: cavity size invariably increases when a restoration is replaced, while the justification for replacement is commonly a marginal morphological discrepancy rather than a diagnosis of secondary caries. The paper supplies a modality-free diagnostic, a coverage reading independent of teacher dosage, a depletion series, three tests on which it stands or falls, and a retrospective-check protocol.
occupation-by-answer; alternative-route coverage; autophagic structure; order of arrival; remainder depletion; additive ledger
The Holding Position: Withdrawal Loss Is Set by Type, Not by Frequency
Coupling an old distinction to an old dependent variable, and the three ways that coupling can fail
A student writing in a foreign language faces a set of irreducible uncertainties. The question is who holds them while the work is being done. On “external provision displaces the learner’s own judgement,” this paper is not the first arrival: truths can be imparted, procedures can only be inculcated is Ryle, 1949; tacit knowledge is Polanyi, 1958; the separation of conception from execution is Braverman, 1974; the ironies of automation are Bainbridge, 1983; and that preset criteria cannot transfer a teacher’s tacit quality concept to students is Sadler, 1989 — restated in 2009 as the indeterminacy in the use of preset criteria. On “withdrawal of external support degrades performance,” withdrawal experiments date to 1959. On the distinction between what can be checked and what can only be accepted, the epistemology of testimony has debated reductionism against anti-reductionism for four decades, and human-automation research names it appropriate reliance. The paper claims none of these. It claims one coupling: the size of the withdrawal loss is set by the share of support of the non-self-checkable type, and not by the frequency, quantity, or cognitive effort saved. The guidance hypothesis has predicted withdrawal loss from frequency for forty years; this paper predicts it from type, and the two yield incompatible experiments.
holding position; non-self-checkable share; withdrawal loss; type versus frequency; nominally checkable; appropriate reliance
The Stopping Point: A Stopping Rule Without a Held-Out Set
On the homology between an instructional termination criterion and the process it judges
Every instructional intervention must stop, yet none of the four ledgers in current use records the basis on which it stopped. In practice, termination is triggered by the learner’s immediate response: a nod, a correct answer, the absence of further questions. On the unreliability of that signal this paper is not the first arrival. Within second language acquisition the signal already has a name — uptake — whose validity has been contested for nearly thirty years (Mackey & Philp 1998); the learning/performance distinction has an authoritative review (Soderstrom & Bjork 2015); the orthogonality of criterion and sensitivity is the first lesson of signal detection theory (Green & Swets 1966); and “measurement changes what it measures” has been named independently at least four times, the earliest monograph appearing in 1966 (Webb et al.). This paper therefore claims none of these. It claims the position they jointly leave empty: an instructional stopping criterion takes its value from inside the very process it judges — it is a stopping rule with no held-out set. The homonymous “early stopping” of machine learning is a deliberate regularizer and a good thing precisely because its criterion is read off a validation set held out from training; classroom early stopping is harmful not because it is early but because its criterion is homologous with the quantity being optimized. The overturning material comes from clinical trial methodology itself: trials stopped early for benefit report larger effects, and that difference is independent of the presence of statistical stopping rules (Bassler et al. 2010) — a pre-specified boundary does not remove it.
stopping criterion; homology; held-out set; post-decision survival interval; closed-loop correction; uptake validity
Determination Before Maturity: An Early Verdict Is a Deletion, Not a Misjudgement
Three fields each built a device to preserve the control; education built none
A learner is classified before the relevant ability has matured — streamed, graded, written into a report, judged “not a language person.” The usual worry is that the judgement may be wrong. This paper argues something else: the cost of a determination made before maturity is not that it is made on insufficient information, but that the learner’s mature state no longer exists; hence the error rate of that determination cannot be estimated from any follow-up data, which can only measure how much remains, never what would have been. That early judgements are unreliable, that labels are self-fulfilling, that tracking widens gaps, and that rejected candidates’ outcomes go unobserved are all prior art — 1948, 1970, 1939, 1979, 1986 respectively. But that literature says uniformly that the counterfactual is unobservable; nowhere does it say the counterfactual has ceased to exist. The overturning material comes from endocrinology itself: exogenous administration produces not a masked reading but the atrophy of the corticotrophin cells of the anterior pituitary, and three years after withdrawal insufficiency persists in 15% of retested patients. A recovery test measures what remains, not what would have been. Forestry keeps uncut reference stands; breeding carries unselected populations forward; endocrinology has the stimulation test. Each field built a device to preserve its control. Education built none. The paper therefore offers no error rate, only a countable reversibility rate, and states the three ways its claim can fail.
early determination; deletion versus concealment; reversibility rate; control-preserving device; mature state; inestimability
Where the Cost Lands: The Sign Is Set by the Parity of the Number of Steps, Not by Strength or Distance
A composite of three transfer mechanisms, and the one condition under which it fails
That the cost of a teaching reform leaves the place it was assigned needs no arguing, and this paper does not claim it. That burdens move by structural parameters rather than by statutory assignment is standard tax incidence; that safety improvements are offset by behavioural adjustment is Peltzman, 1975; that purposive action produces unanticipated consequences is Merton, 1936; that evaluating policy from pre-policy behavioural regularities fails systematically because agents re-optimise is the Lucas critique, 1976; that general-equilibrium effects differ from partial-equilibrium estimates by an order of magnitude was shown in 1998; and that a long-run effect may run opposite to the short-run one is what system dynamics calls policy resistance and compensating feedback. The paper claims one thing: the sign of a reform’s effect on a given party is set by the parity of the number of transfer steps between that party and the point of intervention — not by the strength of the reform, and not by how far that party is from it. The reason is that each family’s transfer link reverses sign. The counterintuitive corollary: parties two steps out are hidden beneficiaries, and parties three steps out are harmed again. The claim is overturned by material internal to ecology: in real webs, several paths of different lengths connect two nodes, so the net sign is not set by any one path and is frequently indeterminate. The paper’s countable reading is therefore the odd-path share, which, near one-half, marks the region where the sign is in principle unpredictable.
cost transfer; parity reversal; odd-path share; path inventory; multi-path indeterminacy; reform evaluation
After the Designer Leaves: In the Gap, One Incomplete Restart Is Worse Than Doing Nothing
Preservation and awakening are two different condition sets, and a prediction that collides with the savings effect
What happens to a person’s foreign language after they leave school? This paper first corrects its own earlier premise. The previous draft said freeze — twenty flat years — and hung this on the classic fifty-year retention study. That is a misreading: the study is cross-sectional, one measurement each from people at different intervals, and cannot in principle describe what happened inside any individual. Genuinely longitudinal work reports divergence: some continue to improve, some clearly decline, with individual six-month change ranging from −12% to +21%. The premise is therefore rewritten from freeze to divergence. That life transitions open a window of unusual behavioural plasticity, that repeated failure lowers self-efficacy, that disuse decays skill, that transitions between care settings concentrate harm, and that follow-up interval predicts readmission are all prior art, the earliest from 1977, and none is claimed here. The paper claims one thing: an incomplete restart during the gap is not a neutral event; it moves the ability into a state that is harder to wake when a genuine opportunity later arrives. The decisive variable in the gap is therefore neither a successor nor motivation, but the number of abandoned restarts. The warrant is seed science’s own secondary dormancy. The paper states plainly that this collides with the savings effect in memory research and rests its only empirical foothold on a 2023 demonstration of anti-savings in motor learning.
post-departure gap; secondary dormancy; preservation versus awakening; abandoned restarts; savings effect; divergence
Recruitment Radius: The Net Loss of Reachable Range in Foreign-Language Teachers’ Target Language
A claim about causal direction, and how it differs from complementary distribution, arrested development, and cue mismatch
A foreign-language teacher speaks the target language every working day. What happens to that language after twenty years? None of the four accounting systems around a language lesson has a field for it: the teacher’s ledger books coverage, the assessment ledger books student output, the student’s ledger books personal gain, and classroom-discourse research books turns while coding teacher speech as learner input. Three established accounts answer the question differently — attrition books proficiency against amount of use, fossilization against saturated communicative need, illusions of competence against overly predictable practice — and all three share one assumption: that a person’s reachable range is a quantity belonging to that person, readable off any single dimension. This paper argues otherwise. The difficulty such a teacher meets outside the classroom register is not that the register was never used and therefore never grew (the complementarity account of Grosjean 2016 and the bounded-register account of Freeman et al. 2015); not arrested development at a stable ceiling (Ferguson 1954, 1956; Ericsson & Towne 2010); not a retrieval-cue mismatch that a reinstated context would undo (Tulving & Thomson 1973; Luchins 1942). It is a net loss: continued successful recruitment inside the classroom register costs reachable range outside it, in proportion to the amount of that successful recruitment, and the loss is not reversed by instruction or context reinstatement. The material that overturns the shared assumption comes from immunology itself: serum responses to repeat immunization derive overwhelmingly from primary-cohort B cells, yet this “primary addiction” decreases sharply as a function of antigenic distance (Schiepers et al. 2023) — breadth is not a property of the individual but of individual × exposure history × distance. The paper supplies a modality-free diagnostic, a dimensionless reading, a coding manual, three tests on which it stands or falls, and a locked pre-registration whose most likely outcome will force the paper to withdraw the word “threshold.”
recruitment radius; far-field recruitment share; net loss; setting distance; teacher target language; organized by event